Verify Your Integration

Use this guide before you ship an AnyInt integration. It is not a payment-style sandbox guide. AnyInt is an AI API platform, so verification means proving that your API key, model IDs, request format, streaming path, task polling, callback handling, and retry behavior work in your own environment.

Verification checklist

1. Configure the key safely

Use an environment variable instead of hardcoding the key.

export ANYINT_API_KEY="your-anyint-api-key"

Do not send API keys to browsers unless your product has a deliberate server-side mediation or short-lived-token design. Most production applications should call AnyInt from a backend service.

2. Discover model IDs

Call the Models API before hardcoding model names.

curl https://gateway.api.anyint.ai/openai/v1/models \
  -H "Authorization: Bearer $ANYINT_API_KEY"

Confirm that the response contains the model you plan to use:

{
  "data": [
    {
      "id": "claude-sonnet-4-6",
      "display_name": "Claude Sonnet 4.6"
    }
  ]
}

Use data[].id in later requests. Do not use a marketing label or a model ID copied from a different account without checking availability.

3. Make a minimal request

Start with the OpenAI-compatible Chat Completions API unless you need a provider-native request shape.

curl https://gateway.api.anyint.ai/openai/v1/chat/completions \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "messages": [
      {"role": "user", "content": "Reply with exactly: AnyInt OK"}
    ]
  }'

If this fails, check the model ID, API key, endpoint family, and auth header before trying larger payloads.

4. Verify streaming

If your product streams responses to users, verify that your HTTP client can read server-sent events.

curl https://gateway.api.anyint.ai/openai/v1/chat/completions \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "stream": true,
    "messages": [
      {"role": "user", "content": "Stream three short bullet points about API gateways."}
    ]
  }'

For OpenAI-compatible streaming, expect incremental chunks and a final stream terminator. Do not treat the first chunk as the full answer.

5. Verify provider-native request shapes

Provider-compatible APIs are useful when your application already speaks the provider's native format.

Anthropic-compatible example:

curl https://gateway.api.anyint.ai/anthropic/v1/messages \
  -H "x-api-key: $ANYINT_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 128,
    "messages": [
      {"role": "user", "content": "Explain token counting in one sentence."}
    ]
  }'

Gemini-compatible streaming example:

curl "https://gateway.api.anyint.ai/gemini/v1beta/models/gemini-3-flash-preview:streamGenerateContent?alt=sse" \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [{"text": "Stream a short answer about model routing."}]
      }
    ]
  }'

Do not mix request bodies across families. OpenAI uses messages, Anthropic uses Claude-style messages and content blocks, and Gemini uses contents[].parts[].

6. Verify async media or music tasks

For task-based APIs, store the returned task ID and poll the documented query route until the task succeeds or fails.

curl https://gateway.api.anyint.ai/dashscope/v1/services/aigc/video-generation/video-synthesis \
  -H "Authorization: Bearer $ANYINT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "happyhorse-1.0-t2v",
    "input": {
      "prompt": "A cinematic product reveal with soft camera movement.",
      "img_url": "https://your-domain.com/source-image.png"
    },
    "parameters": {
      "resolution": "720P",
      "duration": 5,
      "audio": false
    }
  }'

Poll with the returned task ID:

curl "https://gateway.api.anyint.ai/dashscope/v1/tasks/$TASK_ID" \
  -H "Authorization: Bearer $ANYINT_API_KEY"

Only mark an asset ready after the result contains a success status and an output URL.

7. Verify callback handling

If an API accepts a callback URL, make the callback handler idempotent.

8. Handle expected errors

Read Errors and Limits for more detail.

Production readiness

Before launch, confirm:

1. Your service stores ANYINT_API_KEY outside source control.
2. Your application discovers or periodically refreshes available model IDs.
3. Your first-choice model and fallback model are both enabled for the account.
4. Streaming clients handle partial chunks, final chunks, and interrupted streams.
5. Async task code stores task IDs and has retry-safe polling.
6. Callback handlers are idempotent and have a polling fallback.
7. User-facing errors distinguish invalid requests, auth problems, quota limits, and temporary failures.
8. Logs do not contain API keys, private prompts, or generated private content unless your users explicitly consent.